[ CLICK HERE & TYPE TITLE OF PAPER ] - … 2012/ISS SESSION 7/Pereira.pdf · hypothesis (Baumol et...
Transcript of [ CLICK HERE & TYPE TITLE OF PAPER ] - … 2012/ISS SESSION 7/Pereira.pdf · hypothesis (Baumol et...
A HISTORY-FRIENDLY MODEL OF THE INTERNET ACCESS MARKET: THE CASE OF BRAZIL
Authors Marcelo de Carvalho Pereira, PhD student at Universidade Estadual de Campinas, [email protected] "[ Click here & type Author 2 Name, Organisation/Affiliation, Email Address]" "[ Click here & type Author 3 Name, Organisation/Affiliation, Email Address]" *underline presenting author’s name(s)
Abstract
The objective of this paper is the study of the dynamics of competition in the access market of the
internet sector, through the application of History-friendly agent-based simulation methodology. The
simulation model is based on neo-Schumpeterian evolutionary theory, as well as on the relevant attributes of
contemporary institutional theory. The focus of research is the analysis of the processes of industry structure
organization and change and their impact on interfirm competitive dynamics.
Internet sector was originated from the confluence of telecommunications and IT sectors, under
intense support from the US government. It became a leading economic sector following the privatization
wave that swept the world in the ‘90s. One key driver of the internet sector has been the intense
technological opportunity. However, competition in the internet access market has proved less intense, in
most countries, than in other technology-driven industries, including different segments of the internet sector
itself. The empirical results arising from the analysis of competition in the access market are not adequately
explained by evolutionary or institutional theories individually, or by traditional industrial organization. Our
hypothesis to explain this situation is that some features of the institutional environment, associated to the
evolutionary underlying forces, were determinant for the dynamics of the competition process. However, this
combination of factors is not usual in the agent-based evolutionary models available, requiring careful
modelling of the institutional features. We suggest that the integration of a dual, coevolutionary theoretical
perspective would allow better consideration of stylized facts resulting from empirical analysis on the sector.
A modelling solution is proposed to answer some of the key questions about the dynamics of
competition in the internet access market. In particular, a History-friendly approach seems to be convenient
for the task, given the availability of historical data and the possibility of using it to help parameter
calibration and results validation. Critical relationships, among equipment suppliers, internet access
providers and end users, were modelled in detail. Technological innovation is driven by capital equipment
suppliers, modelled through a proxy “monolithic vendor” whose offer mimics the expected outcome of
Schumpeterian competition. Such vendor provides access provider firms with networks able to service end
users. Access services offer is modelled in two dimensions: price and quality. End user choice is based on
those dimensions but is also influenced by decisions of other users. Strategic choice of access provider firms
is modelled as a local adaptive learning process, reinforcing the importance of both search procedures and
social networks. Model parameters and initial conditions were calibrated using empirical data as reference
whenever possible. Most data used was gathered from the Brazilian market, which is similar to data sets
coming from other countries. Sensitivity analysis was performed to suggest critical parametric space regions
and counterfactual analysis opportunities.
The results provided by the simulation model validated the theoretical hypotheses proposed. The
systemic competitive mechanisms unveiled by simulation analysis were strongly dependent on institutional
features, as expected. The establishment of social networks – among access providers and end users –
induced some relevant emergent properties in the simulated system, which simultaneously reduced
aggressive competition and reshaped user preferences beyond pure price and quality considerations. Longer
cycles of innovation diffusion – due to consequences of unintended decisions of the state – also played a
relevant role in supporting market concentration. These institutional phenomena were strong enough to
produce results that are significantly different from similar models in technologically dynamic industries but
close to the empirical evidence gathered from the internet access industry. Model results made clear the
importance of a coevolutionary, History-friendly modelling approach to the analysis of industries like
internet access services.
Key Words: History-friendly, agent-based, simulation, industrial economics, evolutionary, institutional,
complexity, internet
Paper
1. Introduction
The internet sector1 is one key element of what some authors call the transition to the “information
economy”. The sector was originated from the revolution of the information and communication
technologies (ICT). The telecommunications industry, a key component of the internet sector from its
inception, became an even stronger driver of development from the 1990s, after the privatisation,
deregulation and competition introduction processes occurred in most countries. In this scenario, superficial
analysis would envisage the resulting internet access services market (IASM) operating under strong
competition, due to the promising association of low barriers to entry, rapidly growing demand and
significant technological opportunities. However, IASM seems better described by low intensity competition
in some countries, like Brazil. The apparent contradiction between an attractive market to innovative entrants
and the reduced competition verified in practice is the central question here.
Empirical research showed that adequate answers for this question require a somewhat deeper than
usual analytical approach, as usual methods like standard industrial organization could provide only partial
answers, at best. We advocate that one critical reason for the observed outcomes is the importance of
institutional phenomena for the competitive dynamics. Furthermore, empirical evidence points also to the
importance of fast evolving technology in the shaping of the internet sector as a whole. In principle, this
suggests a neo-Schumpeterian evolutionary approach as an appropriate way to understand sectoral
competition. Most of the technological innovation in the IASM is embedded in capital equipment, centrally
developed by a few large multinationals and made available to domestic internet access service provider
(IASP) firms operating in markets all over the world. Nonetheless, the competitive configuration of the
IASM seems to be country specific, notwithstanding the technology availability. This suggests that
technology dynamics, despite significant, may have limited potential to explain large asymmetries
experienced between domestic IASMs.
To supplement an evolutionary approach, our key analytical hypothesis is that country-specific
differences are due to the relatively heterogeneous institutional frameworks, to a large extent. An
institutional perspective seems to articulate well with evolutionary theory, given the potential
complementarity between both. An institutional line of inquiry allows for the improved appreciation of inter
and intra-sectoral interactions and the role of relevant social factors, like culture, shared cognitive
frameworks, social networks, power and the state. Nevertheless, as remember Dosi et al. (2005), this
articulation is not without risks. If, on the one hand, it avoids an innocent perspective of technological
determinism, often attributed to Schumpeterian reasoning, on the other hand, it opens space for a radical
form of social constructivism.
With those issues in mind, we propose modelling both the institutional and evolutionary mechanisms
in action by adopting agent-based simulation techniques to investigate the processes that organize
competition. Obviously, one main task of the model is to test how well the institutional dominance
hypothesis holds. From a methodological standpoint, we embrace the History-friendly approach, proposed by
Malerba et al. (1999), as general guidance. On the empirical side, we selected data from Brazil to set up the
model. We believe Brazil is a compelling case to start with, because of the reasonably complex institutional
scenario and the availability of detailed data. From there it should be straightforward to reconfigure the
model to handle conditions applicable to other countries.
Competition is a broad theme, so it is necessary to define our targets clearly. This paper represents
only a first step of research, about presenting the overall and initial results produced by the model. Further
refinement and detailing of the analysis is unquestionably necessary. Here, we propose focusing on the
general processes driving firms’ decisions, in terms of product prices, qualities and quantities, as well users’
preferences and choices and the resulting market organization. Under a somewhat restricted approach,
market organization and competition are evaluated in terms of market share concentration evolution, firm
entry/exit dynamics and services price/quality/margin trajectories.
The paper is organized as follow. Next section presents the literature supporting the theoretical
framework employed. Section three offers an appreciative empirical analysis of the IASM in Brazil,
1 Pertinent segments of the internet sector include: access services, equipment manufacturing, systems development and content
provision.
providing an overview of the key stylized facts identified. Section four presents some key specifications of
the simulation model. In section five, the main model results are analysed, and brief explanations for the
stylized facts are proposed, based on the model’s internal mechanisms. The paper closes with a review of the
main conclusions.
2. Background literature
From the classical economics in the XIX century, market organization and competition have been
influential subjects. As early models of perfect competition, monopoly and standard oligopoly fell short in
providing adequate explanations to the XX century complex oligopolies (see Chandler, 1990), new
theoretical approaches developed from the 1930s. A new field of studies was created, industrial
organization/economics, initially dominated by the structure-conduct-performance paradigm (Bain, 1959)
and, more recently, by the extensive use of game theory (Tirole, 1988).
Industrial organization introduced several new concepts useful for market analysis. The relevance of
(static) barriers to entry (Bain, 1959) is a key concept to understand market situations where entry of new
firms is difficult or unlikely, sometimes leading to natural monopolies, as is the usual explanation for the
telephony monopolies that prevailed until the 1990s. More recent developments, like the contestable markets
hypothesis (Baumol et al., 1982), network effects (Katz and Shapiro, 1985) and the Stackelberg-Spence-Dixit
model (Tirole, 1988), provided further analytical tools. However, notwithstanding some relevant insights on
explaining concentrated markets, mainstream industrial organization failed short so far in handling complex
dynamics comprehensively, as pointed by several authors (Nelson and Winter, 1982; Dosi, 1982; Kirman,
1997; Metcalfe, 1998; Pyka and Fagiolo, 2005).
Based on Schumpeter’s (1943) creative destruction perspective of capitalist interfirm competition,
Nelson and Winter (1982) proposed evolutionary theory. In such perspective, competition is not directly
related to static efficiency because of innovation – technical or organizational – that relentlessly change the
competitive environment, by dynamically redefining the relative advantages hold by competing firms,
making ex ante definition of competition organization impossible (Dosi and Nelson, 2010). Evolutionary
theory is particularly adequate to explain sectors driven by the technological dynamics and the interaction of
the agents beyond pure market transactions (Malerba, 2006). Competing firms have different capabilities
(Teece et al., 1997), on top of what they try to adapt continuously to the competitive scenario by innovating.
Successful innovators grow and eventually increase their profits; others shrink and may get out of the
market. Fitness, in this scenario, represent the skills that bounded rational2 firms have to solve the specific
problems – technical, organizational or political – they face in the competitive selection process (Cyert and
March, 1963; Nelson, 1995).
When Schumpeterian competition takes place, market structure becomes endogenous (Nelson and
Winter, 1982), presenting itself as an emergent property of the differential innovation capabilities among
firms (Metcalfe, 1998). Heterogeneous capabilities represent contradictory forces leading, at the same time,
to oligopolistic markets and to turbulent competitive dynamics, being the sector-specific balance between
both determinants to industry organization (Dosi, 1982). However, the coexistence, within the same sector,
of markets under highly distinct competitive profiles is not straightforward to grasp from a pure evolutionary
analytical standpoint. Once those markets share the same technological regime3 (Malerba and Orsenigo,
2000), some similarities would be expected, as suggested by the typologies proposed by Schumpeter (1942),
Pavitt (1984), Breschi et al. (2000) or Klepper (2006): high turbulence (intense entry and exit), frequent
technological innovation and constant erosion of incumbents’ dominance and market shares.
The application of concepts derived from institutional theory, in particular the approach proposed by
the organizational studies (DiMaggio and Powell, 1983), seem able to clarify some points not addressed by
evolutionary theory. Under the approach originated from organizational studies, institutions have to be
considered beyond their normative and regulatory aspects, by including a cultural-cognitive instance
(DiMaggio, 1988; Powell, 1991). In this perspective, cognitive structures shared among actors are also
institutions, because they condition the behavioural alternatives available to agents (Scott, 2008).
2 Bounded rationality is a residual category proposed by Simon (1979), characterized by any form of rationality inferior to
omniscience, or substantive rationality, due to cognitive limitations of individuals under a strong form of uncertainty. 3 As defined by the relevant technological features, like the available opportunities, the appropriability conditions, the knowledge
cumulativeness profile and the nature of the knowledge base (Pereira, 2012).
Institutionalization, in such context, is the process where patterns of behaviour or thought become shared by
actors (Jepperson, 1991; Dequech, 2009). In addition to the instrumental and formal institutions considered
by new institutional economics authors (see North, 1990; Williamson, 2000), the organizational studies
approach gives analytical emphasis to the roles of culture, cognition and social interaction in producing
informal and taken-for-granted types of institutions (DiMaggio, 1988; Thornton and Ocasio, 2008). Because
cultural-cognitive elements are based on preconscious, taken-for-granted premises, they constitute the deeper
level of the institutional framework (Beckert, 1999) and so cannot be assumed only as an instrumental tool
created by agents (Battilana et al., 2009).
Cultural-cognitive, normative and regulatory elements are the constitutive blocks of institutions
(Scott, 2008), and their alignment is critical to institutional persistence (Tolbert and Zucker, 1996). Those
elements, when misaligned, represent a resource to agents willing to change the institutional framework for
its own purposes, in what is frequently called institutional entrepreneurship (DiMaggio, 1988; Garud and
Karnøe, 2001). Culture and mental models provide the cognitive elements agents require to provide sense to
the actions of other individuals with whom they interact (DiMaggio and Powell, 1983) as well to perceive
the prevailing institutions and their changes (Denzau and North, 1994; Dobbin, 2004). As a result, agents
adopt shared mental models to structure their action and interaction, while taking in account their objectives
too. The existence of taken-for-granted institutions is not disconnected from purposeful action (DiMaggio
and Powell, 1983). A common cultural-cognitive context is required to enable interaction within a field, thus
associating an institutional framework with the context of the social interactions in that field (Bourdieu,
1972). Fields are made of specific social networks, and these networks generate differentiated power
positions to be fulfilled by agents (Hardy and Maguire, 2008; Beckert, 2010). This implies the consideration
of power relations in the establishment of cognitive structures that are the foundation of taken-for-granted
institutions (Dobbin, 2004). As a consequence, it is expected that different social network arrangements
produce distinct field organizations and institutions (Fligstein, 2001b).
New institutions depend on agents with adequate social skills, able to introduce new ideas and
meanings in their networks of influence and induce cooperation and accommodation between potentially
competing groups (Fligstein, 2001b). Ideas – mental schemes or premises – are powerful tools to the
institutionalization process because they provide actors with cognitive frames that justify and legitimate
action (Scott, 2008). Institutionalization and legitimation are critical steps of institutions development,
allowing their gradual transition from conscious habituation to cultural objectification, when social
consensus is achieved (Tolbert and Zucker, 1996). However, this development is not automatic; conflicts,
contradictions, and ambiguities are intrinsic to the process (DiMaggio and Powell, 1991). Failure in
conciliating interests or identities may block institutional consolidation or accelerate its decline (Fligstein,
2001b), once there is no structural “guarantee” of permanence (Storper and Salais, 1997).
In a perspective of “unstable” institutions, relying on social networks, powerful actors’ role becomes
relevant, because these players depend on the stabilization of institutions to keep their power (Thornton and
Ocasio, 2008). Consequently, incumbents have a common interest to minimize the impacts of challenger’s
actions and avoid institutional entrepreneurship (Fligstein, 1997; Hardy and Maguire, 2008). While more
frequently restrictive, under certain conditions institutions enable skilled challengers of the existing
institutional order (DiMaggio, 1988; Hwang and Powell, 2005). Formal or tacit agreements between capable
incumbents are a strong form of collective action to allow for a stable order under their control, being
stabilization and reproduction of fields crucially dependent on the social skills of these players (Giddens,
1984; Powell, 1991).
When markets are analysed as organizational fields (Bourdieu, 1972), it becomes evident that
hierarchical networks may foster specific taken-for-granted market institutions, usually aligned with the
interests of incumbents (Fligstein, 2001a). Field theory, then, helps us to understand how heterogeneity,
conflicts and strategic actions of agents may be reconciled with stable markets, as more frequently observed
(Powell, 1991; Fligstein, 1997). Economic processes are simultaneously “constrained and carried by
networks defined by recurring patterns of interaction among agents” (Arthur et al., 1997:6). By “absorbing”
individual agents, “social networks are the carriers of new economic practices and new ideas of what it
means to be rational and efficient” (Dobbin, 2004:5), relativizing the role of agency in its stronger forms.
Market development is in part a product of historically created institutional and political
arrangements (North, 1990; Storper and Salais, 1997). The appearance of formal and informal governance
structures, that regulate cooperation and competition in a sector, is the outcome of active institutional
entrepreneurship during its emergence or transformation (Coriat and Weinstein, 2005). In this perspective,
the neoclassical price competition mechanism becomes representative of the failure of the governance
construction process, where the absence of coordination among agents turns aggressive price competition the
de facto mode of governance (Powell, 1991). On the other hand, successful governance institutionalization
may help reducing price aggressiveness of firms and easing market stabilization (Fligstein, 2001a).
In summary, the proposed framework for sectoral analysis is based on the premise of dual dynamics,
where technological and institutional vectors drive the competition organization. Both vectors have
evolutionary nature, in the sense they involve trials, errors and learning along path dependent trajectories in
the historic time (Nelson and Winter, 1982; Storper and Salais, 1997). This coevolutionary scheme is
suggested by authors from both traditions (Hodgson, 1988; Nelson and Sampat, 2001; Fligstein and Dauter,
2007; Scott, 2008) as a more comprehensive analytical perspective in certain scenarios.
The adoption of simulation models, as analytical devices, is a feature of evolutionary theory from its
inception (Nelson and Winter, 1982; Garavaglia, 2010). History-friendly models represent a second
generation of evolutionary models, focused on the study of specific industrial sectors and their time
trajectories, at a more limited level of generality (Pyka and Fagiolo, 2005). On the other hand, simulation
usage is less frequent in institutional studies, despite several recent advances (Arthur, 2000). Complex
economic systems are usually better modelled by agent-based simulation, notwithstanding the incipient
methodological standardization, in comparison to other alternatives (see Metcalfe and Foster, 2004;
Tesfatsion, 2006). The complexity perspective, which largely backs agent-based modelling, privileges the
inquiry on “meso”-level phenomena, essential to represent the heterogeneous networks of social
relationships present on real markets (Potts, 2000; Colander, 2005). Dynamic processes determining network
connections, under the strategic action of agents and the institutional environment they are subject to, lead to
emergent events better understood under this perspective (Holland, 1988).
3. Appreciative empirical analysis4
Borrowing the fortunate concept of Breschi and Malerba (1997), we suggest that a sectoral system of
innovation and production perspective of the internet is an adequate approach to the sectoral appreciative
analysis (Edquist, 2004). In this line, analysis should focus around a given group of close products, in our
case the internet access services, and empirical investigation shall try to explore the relationships among
agents (supply and demand), knowledge (including technologies) and institutions (Malerba, 2005).
A prominent characteristic of the internet is its governance organizations. Derived from innovative
institutional entrepreneurship, internet’s regulation and standardization bodies are powerful, mostly non-
governmental and open to most of the sectoral actors (Mowery and Simcoe, 2002). These organizations
were crucial to the required coordination of agents, in a complex and uncertain technological environment,
leading to the construction of highly sophisticated knowledge and production networks (Kavassalis et al.,
1996). Because of this institutional setup, technical innovation processes in the sector, to a large extent, took
a collective prospect that defined key properties of its knowledge base (Cerf et al., 2000). Despite the
intrinsic cumulativeness of the knowledge base required to implement the physical internet and the services
around it, the collective dimension of its construction – associated with explicit (non-tacit) standardization –
resulted in relative low levels of appropriability. This potent mechanism offered vast technological
opportunities for an unusual large number of competing agents; collaboration and competition processes
were key factors to the fast development of the internet and its supporting technologies (Corrocher, 2001).
Simultaneously with internet development, the telecommunications sector went through a significant
change process during the 1990s. State owned monopolies all over the world were swiftly privatized,
frequently at the same time when competition was introduced in those markets (Edquist, 2004). Considering
the growing importance of data communication infrastructure to the deployment of the internet, telecom
operators’ legacy physical networks naturally became the initial fabric of the internet (Dalum and Villumsen,
2003). Remarkably, physical networks were not the only legacy from the telecom sector to the internet.
Taking advantage of their early hold of essential parts of the new sectoral system, privatized telecom
operators typically leveraged their position in the florescent IASM (Davies, 1996), obtaining significant
competitive advantage during the critical market formation period (Edquist, 2004). This experience is
4 For details on all empirical data presented in this section, and the respective sources, see Pereira, 2012.
markedly different from other segments of the internet sector, like hardware, software or content, where
many prominent firms are relatively young, originated inside or around this other growing internet markets.
It should be noted that IASM concentration has some distinctive characteristics from the situation
prevailing during old-time telephony monopolies. Owing to the mostly nonproprietary and non-tacit nature
of internet knowledge base and technologies, as well the strong standardization efforts of its governance
organizations, interconnection among competing physical networks are almost universal and costless5.
Consequently, the significant network externalities offered by the internet do not provide larger IASPs with
relevant competitive advantages in most situations6. Thus, differently from the natural monopoly case of
telephony, it is in principle possible to entrant IASPs to challenge incumbent operators successfully (Noam,
1994), as demonstrated by relevant examples in many countries.
There are three milestones in access services technological trajectory. Dial-up access was the initial
“narrowband” technology available to most internet users during the 1990s. Based on direct overlay usage of
the existing telephony network, its implementation was painless and not dependent on collaboration from
telecom operators to a large extent. Not surprisingly, this was the most competitive phase of the IASM in
many countries. Fixed broadband was the second mainstream technological step, introduced in the late
1990s. Contrary to dial-up, fixed broadband technologies where specifically designed to take advantage of
incumbents’ network infrastructure, making their offer hard to be replicated by entrants without explicit
support from legacy telecom operators. Mobile broadband is the latest form of internet access, based on the
utilization of state-allocated radio spectrum to provide wireless services. Despite existing infrastructure and
user base provide an edge to incumbent telecom operators, wireless technologies open more competitive
opportunities for entrants. There are other niche access technologies available, like satellite and fibre optics,
but they still have relatively small penetration in most countries.
In most OECD countries, each consecutive technological step has diffused in decreasing timeframes.
While dial-up access took 5 years from the launch of first offers to the beginning of massive adoption, fixed
broadband poured out in less than 4 years, and mobile broadband took between 3 (3G) to 2 (4G) years to
achieve mainstream market penetration. On the other hand, in cases like Brazil the same process has taken
the opposite direction. From 5 years to dial-up diffusion, diffusion took 6 years for fixed broadband and 8
years for 3G mobile broadband (no 4G yet). As mentioned before, new access technologies are embedded in
new network equipment and terminals, developed and supplied by few large multinational firms, so their
availability is relatively universal. Thus, new technology diffusion is critically connected to domestic
operators decisions, sometimes also associated with the availability of some required complementary assets,
like state-granted radio spectrum.
Anecdotal evidence shows that the singular diffusion pattern observed in Brazil was not fortuitous.
Intense action by the incumbents was targeted on the regulatory agency to postpone the issue of licenses to
new operators. The intimate cultural-cognitive and personal connections between operators’ representatives
and the new administration agents, usually coming from monopoly period, enabled the establishment of a
harsh regulatory environment to entrant operators. This effectively prevented any new competitor to
anticipate the launch of both fixed and mobile broadband. When new technologies where finally “allowed”,
the incumbents were in the right timing to embrace them. This was even more apparent in the case of mobile
internet, where 3G/UMTS and 4G/LTE radio spectrum auctions were kept on hold for more than 4 years,
based on the general understanding – between incumbents and administration – that was necessary to
depreciate the existing networks adequately before introducing new technologies. In our view, this was not a
usual situation of regulatory capture. Even market analysts and specialized journalists, at the time, used that
same argument to justify the regulatory agency “moderation”. No relevant debates took place at the time on
the subject of the eventual consequences to the competition (one of the three pillars of the formal regulatory
regime). After all, this was the way telecom infrastructure was operated in the past 100+ years. Under the
state monopoly regime, it was perfectly rational to maximize the lifetime of scarce capital. Thus, it seems
reasonable to suggest, ex post, that this same worldview was taken-for-granted as part the now prevailing
institutional framework.
5 At least among same tier IASPs, but anecdotal evidence is that, even for smaller players, interconnection costs are not significant
barriers for domestic competition in most countries. 6 To most users it is irrelevant if accessing the internet from a large or small IASP, assuming both adopt the same technical quality
parameters, once all networks are interconnected.
Internet access services became a highly concentrated business in Brazil. The 4 incumbent players,
originated from the privatization of the telecommunications monopoly, dominate almost 80% of the national
IASM. If we exclude dial-up access, their share goes over 90%. All the usual indicators (HHI > 0.25, C4 >
0.85) point to high market concentration at the national level, in a scenario of market share stability and
limited competition among the incumbents, which usually concentrated in different geographical regions.
When analysed at the state level7, concentration is even higher: the local privatized incumbent operator alone
holds in average 60% of market share. Despite the formally open market and the 1900+ small firms
providing internet access services in Brazil (as of March 2011), only one new company successfully
managed to enter the IASM and become a significant player, in the last 10 years.
The impact of IASM organization on prices is evident. Minimum access prices are 32% higher than
the OECD average, despite the much lower average access bandwidth in Brazil. The 2011 International
Telecommunications Union (ITU) broadband costs ranking shows Brazilian fixed broadband in position 56
among 165 countries (higher ranks represent more expensive services) and mobile broadband took the last
position among the 21 countries considered. In a similar survey done by UNCTAD, mobile broadband prices
in Brazil got the worst place among 78 countries.
Notwithstanding the high relative prices of internet access in Brazil, average price per connected
internet user has fallen at a fairly steady rate of 13% every year (2004-2010). However, markups were kept
by incumbents at a relatively stable level, above 50%. When comparing price reductions with service
penetration rise, it seems sensible to suppose that it was the interest of incumbents in augmenting the user
base that led price adjustments. The observed price reductions were close to the ones required to match the
increase in the number of users in the period 2005-2010, except for 2006. In other words, if price reductions
were smaller, as was the case in 2006, service penetration growth would have been reduced significantly8.
Consequently, it seems logical to the incumbents to move prices along the demand curve, increasing
marginal revenues while average unit costs keep at least constant9, as for the classical monopolist.
In summary, the evidence gathered in empirical research can be synthesised in four key stylized
facts. First, Brazilian IASM presents persistent market concentration, with the dominance of legacy
incumbents – originated from the privatized public monopoly – and restricted room for new competitors.
Second, empirical data shows a low rate of successful entry, despite entry is neither formally blocked nor
impossible de facto, as the single counterexample available demonstrates. Third, longer than expected
technological diffusion cycles have characterized the introduction of new generation services, even though
the required technical artefacts were readily available to both incumbents and entrants, in a pattern of
succeeding longer diffusion cycles when compared to more competitive markets. Fourth, evidence suggests
reduced price-based competition, in the face of limited product differentiation and high markups enacted by
existing IASPs, being price reduction apparently instrumental to the growth of incumbents.
4. Model definition
The next analytical step is the specification of the simulation model. The main objective of a
History-friendly model is to test if its theoretical hypotheses are logically compatible and to what extent with
the empirical stylized facts (Malerba et al., 1999; Windrum et al., 2007). However, the purpose of the model
goes beyond hypothesis testing. It intends, in more general terms, to select, submit, and combine ideas and
hypotheses – including causal relations between variables – while staying compatible with stylized facts
(Pyka and Fagiolo, 2005). Furthermore, the model may generate results that are not immediately or readily
derived from theory, enabling deeper understanding of fundamental causal mechanisms of complex systems
(Axelrod and Tesfatsion, 2006).
The proposed model was specified to study the interactions among sets of users, IASP firms,
technologies and critical institutions, as pointed by theoretical and empirical analysis, to enable the
identification of the main features of an artificial representation of the IASM. The model is based on a set of
difference equations, defining discrete time series for selected state variables of the model. Each simulation
7 Brazil is divided in 26 states plus the federal capital district. 8 On the other hand, higher growths would have been limited by the rise of the number of available unconnected terminals, which are
the upper hard constraint for internet services penetration. 9 There is strong empirical support to the presence of economies of scale on the provision of telecom services, including internet
access.
run is then defined by a set of times series from all state variables. The model is time driven and all
contemporaneous events are supposed to take place simultaneously, in each time step t (t = 1, 2, 3, …, N,
where N is the simulation length). Such contemporaneous time convergence requires that the order of
equations valuation to be specified properly, to avoid ambiguities. This is achieved through the careful
specification of the lag structure of each variable and the definition of a fixed evaluation order for the
equation set.
Behavioural difference equations are processed in the following sequence: (a) the proxy monolithic
network equipment vendor performs technology search, trying to increase the productivity of existing
technology vintages and, eventually, launch new, more productive ones; (b) prospective entrant IASPs
evaluate convenience (profitability and opportunity) of entry, and, if so, select initial network capacity and
strategy; (c) IASPs select prices and investments for the period, given the (myopic) expectations of increase
(or decrease) in the number of users; (d) new potential users come to the market while market saturation is
not reached; (e) users whom do not have an assigned IASP (new comers or without contract) choose a new
IASP, according to their preferences, budget and the influence of other users; (f) IASPs decide about
investment financing and use of profits; and (g) bankrupt or too small IASPs leave the market.
Full model documentation is available at http://sites.google.com/site/modelosetorinternet. Model
specification was coded in C++ using the Laboratory for Simulation Development (LSD) created by Marco
Valente (2002). The model is composed of 42 main equations, of which 25 are critical, because of the
incorporation of key theoretical premises. Full explanation for each behavioural difference equation in the
model is presented in Pereira (2012). Next, we briefly introduce some equations that model critical features
of the model in three areas: demand and supply clearing, strategic learning, and technological innovation.
Demand is modelled through heterogeneous users, in two dimensions: budget and preferences. User
k is interested in contract internet access services for multiple time steps paying a fixed price each. IASP i
offers a single combination of access price and quality in any given time step t. Every time a user is
out of a contract, she ranks all IASPs according to a utility function10
and selects the IASP with the
highest utility considering her budget .
(1)
This mechanism represents an implicit replicator equation (Metcalfe, 1998) because, as all
individual users chose their IASPs, it defines the resulting market shares for each IASP, in every period.
Parameters , and represent user preferences weight in terms of price, quality and market share, for
any IASP. is the weighted market price. is the quality of IASP i as perceived by user k and is its
market share. The third term in (1) is a proxy to the relational influence of other users’ choices on the
individual preferences and represents a positive externality to (larger) firms. It should be noted that this is not
the classical network externality (Shy, 2001), bearing in mind users have no direct benefit in choosing the
same IASP as her acquaintances. On the contrary, such a disposition might cause the user to choose an IASP
with inferior objective attributes (in price or quality), but more “popular”, even in the absence of tangible
benefits.
The quality offered by the IASP to all its users in t, is inversely proportional to the utilization of
its network total installed capacity . By definition, the capital equipment vendor designs one unit of
network physical capacity in order to meet the demand from one user. Thus, is the current number of
users of IASP i. q is a fixed parameter and accounts for nonlinearity between capacity mismatch and quality.
(2)
10 The use of continuous utility functions is criticized for its poor adherence to the empirical experience (VALENTE, 2009).
Nevertheless, the simplicity of a traditional Cobb-Douglas function was preferred for the initial analytical stage.
The total installed capacity depends on the productivity and the stock of each technology
vintage j installed in IASP’s network. Every IASP has distinct vintages in operation at time t.
(3)
IASPs assess the need for increasing installed capacity each time step. All required investment
adopts the most current technology . Firms decide investment based on the expected network capacity
required plus the incurred depreciation . is the unit price of technology . Investment is
subject to a technology-specific minimum scale . is a non-fixed parameter defining the target
quality, according to the current strategy of IASP i.
(4)
Firms plan network capacity prospectively for n periods, by setting expectations for
acquisition (or loss) of new users. Smaller firms (market share below the parameter ) project demand
from the customer base evolution in previous planning period. Parameter represents the qualitative
expectations about the future ( representing accelerating growth). Larger firms ( ) evaluate
future demand in terms of total market growth ( ) and on the expectation of relative
performance ( pointing to market share rise).
(5)
When firm has an expectation of reduction in the number of customers, it keeps the existing installed
capacity. If necessary, reduction of capacity occurs through depreciation without equipment replacement.
Prices are determined, in principle, based on the desired price compatible with target
profitability . However, each available strategy, to be presented next, defines also different
complementary objectives and those may conflict with . is the total running cost per period.
(6)
For example, we present the price setting criterion for strategy type 1 (Table 1). Here, strategy gives
priority to increase market share, while preventing prices below expected unit cost or above . New
price decision is taken in consideration of the rate of market share change and its significance with
respect to the sensibility threshold parameter . is a price change “aggressiveness” parameter.
(7)
Organizational innovation is modelled as an evolutionary process of strategic search by IASPs that
seek “satisficing” rates of return on investment under the largest market share compatible with this rate. To
pursue it, they can adjust their short term goals for price and quality and some other behavioural parameters.
The model allows different algorithms to implement strategies, including adaptive mechanisms, i.e., the
search for better strategies if current strategy fails. This process is based on the comparison of IASP own
results with those of competitors. Thus, the model allows strategies to pass through a selection mechanism
based on learning and imitation. However, the model supports only a predefined set of strategies, listed in
Table 1. IASPs in distinct social groups – incumbents or entrants – have somewhat distinct strategy sets, in a
“small world” organization (Watts, 1999). It is supposed that every firm knows the set of strategic
alternatives available in its social group and their average performance over time.
Table 1 – Available business strategies.
ID Strategy Group Description
1 Share seeker I/E Maximize market share, keep profitability at target if possible under fixed quality target
2 Share seeker +
low quality I/E
Maximize market share, keep profitability at target if possible under low quality target
3 Group price follower
I Follow weighted average incumbent price under fixed quality target
E Follow weighted average entrant price under fixed quality target
4 Group price & quality
follower
I Follow weighted average incumbent price and quality
E Follow weighted average entrant price and quality
5 Market price follower I/E Follow weighted average market price under fixed quality target
6 Market price & quality
follower I/E Follow weighted average market price and quality
7 Profit seeker I/E Seek profitability only under fixed quality target
8 Top quality I/E High quality target under high price
9 Low price I/E Set price to weighted average market unit cost under low quality target
(I/E: available for incumbents and entrants; I: incumbents only; E: entrants only)
Strategic learning algorithm works as follow: after a fixed period since the last change of strategy,
each IASP assess whether its profitability target is reached. If so, it maintains the current strategy. If not,
it evaluates whether the strategies of competitors, within its social group, are providing better outcomes over
time, in terms of the weighted average results obtained by the adopters. If this is the case, it imitates the best
strategy. In exceptional situations (multiple periods of negative cash flow or market share close to zero), the
assessment of strategies becomes less demanding and imitation requires only profitability or market share
exceeding those of problematic IASP. Entrant IASPs pick the best current strategy practice on start.
There are two types of technological innovation in the model: incremental, associated to
improvements of existing technology vintages, and radical, when new vintages are introduced by the proxy
monolithic vendor. Accordingly, two types of search routines are configured, both modelled as two-step
stochastic, productivity-enhancer processes. At any time, there is a single best practice in terms of the most
productive technology and all IASPs are aware of it. Thus, stochastic components are not present in the
technical search of IASPs, since the model assumes that they simply pick the most current equipment
available when required.
There is a probability in each time step of a technological advance. This probability
has Poisson distribution as presented and (incremental innovation of existing vintages) or (radical
innovation, generating new technology vintage) is the success parameter.
(8)
If first stage spawns an advance, a new potential for productivity (incremental) or (radical)
is produced with normal distribution, based on current productivity or respectively. Standard
deviation of incremental productivity improvement is decreasing as technology gets older. ,
and are parameters.
(9)
(10)
Technological advance is adopted only if it improves productivity.
(11)
5. Model main results
Most of the model’s 41 parameters and 9 lagged variables requiring non-trivial initial conditions
were calibrated using empirical data, as appropriate in a History-friendly approach. Simulation time was
adjusted so 1 time step is equivalent to 1 quarter (3 months). All model results were evaluated by statistical
parameter estimation over samples of 100 simulation runs, due to the presence of stochastic elements in the
model. Sample size was selected to ensure at least ±5% precision at 95% confidence level. Statistical
distributions of most variables were unimodal and sufficiently symmetrical to justify the adoption of
averages and standard deviations as representative parameters of model results. After initial calibration,
sensitivity analysis of all parameters and initial conditions was performed, to identify critical parameters.
Parameters and initial conditions were extensively tested around calibration figures in ranges large enough to
encompass maximum and minimum values compatible with reasonably expected empirical magnitudes.
Interestingly, only a relative small number of parameters were critical on producing the main model results
associated to the simulated market organization. Impact analysis of parameters and initial conditions on 10
selected structural indicators11
was performed by ANOVA tests at 1% significance. Of 50 parameters and
initial conditions, 13 showed overall significant statistical impact, but only 5 were relevant in a qualitative
dimension, meaning their variation generated different competitive outcomes effectively. For details on each
step of model setup and test, see Pereira (2012).
Simulated IASM starts with 4 IASPs and 1.8 million potential service users, conforming to empirical
data. Potential user growth is modelled as a contagion process, leading to the usual logistic shape, adjusted to
the Brazilian data. User growth reaches saturation around . New users have random individual
budgets distributed according to real data. They also have heterogeneous preferences defined randomly and
uniformly over the allowed ranges.
Observing model output, it is not evident that price-based competition is moderate. Weighted
average price and quality in the virtual market had an undeniable down trend, more intense for prices. During
the phase of fast market growth ( average prices fall quickly, but stabilize afterwards, as detailed in
Figure 1 (cf. calibration curve). Conversely, average profitability decreases during the fast growing phase
11 Indicators included concentration indexes, number of firms, market size, profitability, age of competitors, and market price and
quality weighted averages and variances.
and stagnates after all, as represented by the gap between the average price and unit cost in Figure 1.
Nonetheless, incumbent’s rate of return on invested capital (RoIC) can be up to 10 times higher than
entrants’ when market matures ( ). These model outcomes are all compatible with empirical data.
Figure 1 – Average weighted price and unit cost per time step (in BRL).
(Empirical calibration plus 2 price sensitivity scenarios).
Analysis of model runs shows that incumbents usually decrease prices less frequently than entrants,
due to the stronger “lock-in” of users to larger IASPs and, more important, to the typical strategic profile
adopted by them. During the growth phase, entrants drive price-based competition, by usually choosing
strategies more aggressive than incumbents. Strategies 1, 2 and 7 (see Table 1) typically predominate among
incumbents, representing a higher priority on profit-oriented targets (preventing price reductions whenever
possible). On the other hand, entrants more frequently adopt pure price strategies (3, 5 and 9), becoming
more frequently involved in aggressive price-based competition led by survival pressure. IASPs are free to
select strategies from the available options, by an adaptive process where local learning is critical for the
results. When compared to a counterfactual scenario where there is no learning from the choices of others,
the calibration scenario represents a remarkable reduction in price-based competition and slower price
erosion. Under the adaptive-without-social-learning counterfactual scenario, virtual market dynamics and
organization get closer to standard Schumpeterian competition: incumbent average lifespan is reduced
significantly, entry becomes less risky and concentration gets weaker (but far from perfect competition). As a
consequence, in the counterfactual scenario, price and margin erosion is more intense.
The strategic divergence between incumbents and entrants, to distinct profiles, creates an intriguing
emergent phenomenon that reduces dominant players’ aggressiveness and helps the stabilization of the
market. This seems in line with empirical evidence, supporting the stylized fact that reduced price-based
competition is a characteristic of the IASM, compatible with the selected theoretical framework. However,
differentiated strategic profiles are not the only mechanism preventing more accelerated prices decline.
Surprisingly, another relevant element is the threshold that defines the minimum price changes considered by
users (pstep, set at 5% in model calibration). The threshold models a known cognitive characteristic of users,
whom usually do not acknowledge price differences they subjectively consider as “too small”. This
characteristic is recurrently mentioned on anecdotal evidence on the firm’s price setting routines. Figure 1
presents the impact of different pstep threshold levels in overall average prices (cf. the pstep indicated curves).
If IASPs could perform smaller adjustments in their prices during the competitive process, model shows that
price erosion may slow down significantly and conversely. However, this depends entirely on modifications
in cognitive frames usually shared among users, in the real system, and so on a form of institutional change.
Obviously, it is in IASP’s interest to be able to adjust prices in the smallest possible steps (limited only by
“menu costs”), as to minimize the impact of a price reduction. However, it is not so obvious that this option,
globally, would create an emergent form of price rigidity that is negative to users. Said in another way, the
more the users become collectively sensitive to price changes, the smaller are going to be the price
reductions due to the competitive process.
0
50
100
150
200
250
300
350
1 26 51 76 101 126 151 176 201 226
Avg. unit cost
Avg. price (Pstep = 20%)
Avg. price (calibration)
Avg. price (Pstep = 1%)
The total number of IASPs in the market usually grows up to , from 4 to around 10 players,
falling from there on and converging to about 5 firms at (the end of simulation process).
Nonetheless, there is a turbulent process of entries and exits of IASP firms behind those average figures. The
persistent low margins captured by the average entrant make them financially fragile, particularly in
moments of radical innovations, as model produced data shows. The comparatively low RoIC of entrants are
somewhat intriguing, given the usual advantage of more up-to-date technology hold by entrants,
consequently operating under higher productivity and lower costs when compared to incumbents. The
simulation data on average age of networks of incumbents and entrants in Figure 2 (cf. calibration curves)
shows the relevant advantage of entrants.
Figure 2 – Weighted average age of network equipment (in time steps).
Model data analysis shows that lock-in of most users on incumbents’ networks, the presence of
economies of scale, and the low aggressiveness among incumbents are the key drivers of the low RoIC of
entrants and, accordingly, of their higher probability of failure. Moreover, calibration scenario does not
presume any correlation between investments from incumbents and new technology introduction. However,
if we consider the operation of a “synchronization” mechanism between these two matters, results can
change substantially. The curves marked “synchro” in Figure 2 show the effects of new network
technologies being delayed and introduced only when most of incumbents’ networks are old enough for
depreciation, as suggested by empirical analysis. In this scenario, it becomes clear that incumbents are much
more responsive in replacing their networks following a radical innovation, while the behaviour of entrants
barely changes. This move reduces considerably the cost advantage hold by entrants, decreasing their
lifespan expectancy by about 45% in regard to the calibration scenario. This last point can be further
reinforced by Figure 3, which shows the quantitative impact of longer average times between new
technology vintages (radical innovations) on entrant average lifespan. As a result, those model outcomes
seem to be compatible with two stylized facts coming from appreciative analysis, the low rate of successful
entry and a longer than expected technological diffusion cycles, with significant impacts on competition.
0
10
20
30
40
1 26 51 76 101 126 151 176 201 226
Incumbents (calibration)
Incumbents (synchro)
Entrants (calibration)
Entrants (synchro)
Figure 3 – Weighted average lifetime of entrants (in time steps).
Detailed investigation leads to the conclusion that restless turbulence among entrants, associated to
relative stability among incumbents, has an unequivocal outcome: the tendency of lasting concentration of
the IASM in the hands of few incumbents. Figure 4 shows the Herfindahl-Hirschman Index (HHI) trend for
market shares (cf. calibration curve). Calculation of the HHI for capital shares (network sizes) provides
similar results. Concentration, in any case, is substantially above the levels that conventionally characterize a
market as highly concentrated.
Figure 4 – Herfindahl-Hirschman Index for market share.
(Empirical calibration plus 3 counterfactual scenarios)
Simulation results, as exposed, are consistent with the persistent market concentration stylized fact.
It should be noted that this market profile is not a structural outcome of the model; adequate counterfactual
parameter sets can generate remarkably distinct competitive results. Figure 4 presents some HHI results
when employing three different counterfactual scenarios, chosen because they represent some compelling
limit cases. Of interest here are scenarios 1 and 2, where concentration is strongly reduced. Reviewing the
processes enabled by these counterfactual parameter settings, three mechanisms seem to explain the effects
observed: (a) the importance of the reference to the choice of other users in the selection of IASP (b3
parameter), while the single most influential factor to the results obtained; (b) the presence of economies of
scale (cs parameter); and (c) the impact of user subjective acuity among objective differences in quality of
IASPs (q parameter).
0
20
40
60
80
100
1 26 51 76 101 126 151 176 201 226
Período de simulação
prad = 12
prad = 20
prad = 28
prad = 48
prad = 36
0
0.2
0.4
0.6
0.8
1.0
1 26 51 76 101 126 151 176 201 226
Scenario 3
Scenario 2
Scenario 1
Calibration
The importance of economies of scale in a sector such as the internet is probably the mechanism
better established in the literature. However, this factor alone is not capable of changing the model results in
any qualitative way, in spite of its modest quantitative relevance. Even if we eliminate economies of scale
completely, HHI would be reduced, at most, by less than 0.20 as indicated in counterfactual scenario 3 in
Figure 4. On the other hand, two new phenomena, of eminently institutional nature, seem evident in the case:
the importance of social references in IASP selection, through collective choice feedback, and the relative
nature of quality metrics, based on the cognitive inclination of users in not acting on “too subtle” changes in
grades of service. These are not usual empirical justifications for market concentration, despite being points
spotted by other authors, such as Jonard and Yildizoğlu (1998) and Birke and Swann (2006). It is noteworthy
the significant impact on market concentration that takes place even with small changes of parameters b3 and
q. The introduction of endogenous features in individual preferences formation, even in a small proportion
amongst other factors considered by users in their judgment, caused emergent processes of downward
causation nature (Hodgson and Knudsen, 2004). There is a clear feedback process going on here, between
the emerging structure, represented by the set of cognitive schemas collectively adopted by users and the
individual choice of the IASP by users. This feedback increasingly affected the dynamics of the sectoral
structure over simulated time.
6. Conclusions
The rapid convergence of multiple heterogeneous agents to the internet sector represented a complex
institutional building project. The new institutional environment then developed, equally cooperative and
competitive, was a collective form of mitigation of strong uncertainties, associated with a new environment
like the internet, allowing increased investments and attracting new entrants to the industry. However,
different cooperation-competition profiles among industry segments were established. In Schumpeterian
terms, time trajectories guided markets like equipment, systems and content to a more creative destruction-
type dynamics, while others, like access services, apparently took a creative accumulation path in some
countries. In our perspective, this was due, to a large extent, to the persistence of certain institutional
characteristics of the former telecommunications monopoly regime. This legacy, we argument, facilitated the
dominance of the IASM by firms originated from the privatizations of the 1990s. Such circumstances seem
to fit nicely the case of Brazil, as empirical research presented. Appreciative analysis suggests the description
of the arrangements in the Brazilian IASM by at least four stylized facts: persistent market concentration,
reduced competition through price mechanisms, low rate of successful entry, and longer than expected
technological diffusion cycles.
The proposed History-friendly simulation model produced results that were quite close, in qualitative
terms, to those observed in the actual economic system. Some of the main reasons for market concentration
and limited competition were identified as emergent institutional phenomena. Of interest is the strong impact
of other users’ choices in the setting of user preferences (downward causation) and the effects of relational
networks of firms on adaptive strategic learning and aggressiveness profiles. The model also provided
explanations on some mechanisms softening competition, highlighting the sometimes crucial effects of social
networks, established conventions or cognitive issues of governmental agents. The role of technological
dynamics for the IASM organization was clarified, including how its effects are potentially contradictory,
especially in case of successful institutional entrepreneurship. Counterfactual analysis pointed out that a
significantly less concentrated market structure depends critically on unlikely scenarios, based on a change
of some stable empirical parameters, at least in the short term. The dominance of institutional processes does
not mean that traditional elements of industrial analysis, such as those from industrial organization or
evolutionary theory, have not played their expected role. However, as the model demonstrated, some of the
results, usually explained exclusively by these traditional elements, depended crucially on the concurrence of
institutional factors. For example, the model rejected the hypothesis that the removal of economies of scale,
in isolation, would be enough to change market concentration in qualitative terms.
Acknowledgments
I wish to thank David Dequech, José Maria Silveira, Marco Valente, Esther Dweck, Paulo
Fracalanza, Mariano Laplane, Thadeu Silva, and André Gimenez for their helpful comments on earlier
versions of this work or otherwise their invaluable support to it. I am responsible for all remaining errors.
The financial support of the Universidade Estadual de Campinas is also gratefully acknowledged. This work
was done with the support of CNPq, Conselho Nacional de Desenvolvimento Científico e Tecnológico –
Brasil.
References
Arthur WB (2000) Cognition: The Black Box of Economics. In Colander D (Ed.) The Complexity Vision and
the Teaching of Economics. Northampton: Edward Elgar.
Arthur WB, Durlauf S, Lane DA (1997) Process and Emergence in the Economy. In Arthur WB, Durlauf S,
Lane DA (Ed.) The Economy as an Evolving Complex System II. Reading: Addison-Wesley.
Axelrod R, Tesfatsion L (2006) A Guide for Newcomers to Agent-Based Modeling in the Social Sciences. In
Tesfatsion L, Judd K (Ed.) Handbook of Computational Economics, Vol. 2: Agent-Based
Computational Economics. Amsterdam: North-Holland.
Bain JS (1959) Industrial organization: a treatise. Greenwich: JAI.
Battilana J, Leca B, Boxenbaum E (2009) How Actors Change Institutions: Towards a Theory of
Institutional Entrepreneurship. The Academy of Management Annals 3(1):65-107.
Baumol WJ, Panzar JC, Willig RD (1982) Contestable markets and the theory of industry structure. San
Diego: Harcourt Brace Jovanovich.
Beckert J (2010) How Do Fields Change? The interrelations of Institutions, Networks, and Cognition in the
Dynamics of Markets. Organization Studies 31(5):605-627.
______ (1999) Agency, Entrepreneurs, and Institutional Change: The Role of Strategic Choice and
Institutionalized Practices in Organizations. Organization Studies 20(5):777-799.
Birke D, Swann P (2006) Network effects and the choice of mobile phone operator. Journal of Evolutionary
Economics 16(1-2):65-84.
Bourdieu P (1972) Esquisse d'une théorie de la pratique. Genève: Droz.
Breschi S, Malerba F (1997) Sectoral systems of innovation: technological regimes, Schumpeterian
dynamics and spatial boundaries. In Edquist C (Ed.) Systems on Innovation. London: Pinter.
Breschi S, Malerba F, Orsenigo L (2000) Technological regimes and Schumpeterian patterns of innovation.
Economic Journal 110(463):388-410.
Cerf, V et al. (2000) Brief History of the Internet. www.internetsociety.org/internet/internet-51/history-
internet/brief-history-internet.
Chandler A (1990) Scale and Scope: The dynamics of industrial capitalism. Cambridge: Harvard Press.
Colander D (2005) The future of economics: the appropriately educated in pursuit of the knowable.
Cambridge Journal of Economics 29(6):927-941.
Coriat B, Weinstein O (2005) The social construction of markets. Issues in Regulation Theory 53:1-4.
Corrocher N (2001) The Internet services industry: Sectoral dynamics of innovation and production and
country-specific trends in Italy and in the UK. www2.cespri.unibocconi.it/essy/wp/corroch.pdf.
Cyert RM, March JG (1963) A behavioral theory of the firm. Cambridge: Blackwell.
Dalum B, Villumsen G (2003) Fixed data communications: challenges for Europe. In Edquist C (Ed.) The
Internet and Mobile Telecommunications System of Innovation: Developments in Equipment, Access
and Content. Cheltenham: Edward Elgar.
Davies A (1996) Innovation in Large Technical Systems: The Case of Telecommunications. Industrial and
Corporate Change 5(4):1143-1180.
Denzau A, North D (1994) Shared Mental Models: Ideologies and Institutions. Kyklos 47(1):3-31.
Dequech D (2009) Institutions, social norms, and decision-theoretic norms. Journal of Economic Behavior
and Organization 72(1):70-78.
______ (2006) The New Institutional Economics and the theory of behaviour under uncertainty. Journal of
Economic Behavior and Organization 59(1):109-131.
DiMaggio PJ (1988) Interest and agency in institutional theory. In Zucker L (Org.) Institutional Patterns and
Organizations. Cambridge: Ballinger.
DiMaggio PJ, Powell WW (1991) Introduction. In Powell W, DiMaggio P (Org.) The New Institutionalism
in Organizational Analysis. Chicago: University of Chicago Press.
______ (1983) The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in
Organizational Fields. American Sociological Review 48(2):147-160.
Dobbin F (2004) The Sociological View of the Economy. In Dobbin F (Org.) The New Economic Sociology:
A Reader. Princeton: Princeton Press.
Dosi, G (1982) Technological paradigms and technological trajectories: A suggested interpretation of the
determinants and directions of technical change. Research Policy 11(3):147-162.
Dosi G, Nelson RR (2010) Technical Change and Industrial Dynamics as Evolutionary Processes. In Hall B,
Rosenberg N (Eds.) Handbook of the Economics of Innovation, Vol. 1. Amsterdam: Elsevier.
Dosi G, Orsenigo L, Labini MS (2005) Technology and the Economy. In Smelser N, Swedberg R (Org.) The
Handbook of Economic Sociology. Princeton: Princeton Press.
Edquist, C (2004) The fixed Internet and mobile telecommunications sectoral system of innovation:
equipment production, access provision and content provision. In: Malerba, F (Ed.) Sectoral Systems
of Innovation: Concepts, Issues and Analyses of Six Major Sectors in Europe. New York: Cambridge
Press.
Fligstein, N (2001a) The architecture of markets. Princeton: Princeton Press.
______ (2001b) Social Skill and the Theory of Fields. Sociological Theory 19(2):105-125.
______ (1997) Social Skill and Institutional Theory. American Behavioral Scientist 40(4):397-405.
Fligstein N, Dauter, L (2007) The Sociology of Markets. Annual Review of Sociology 33:105-128.
Garavaglia, C (2010) Modelling industrial dynamics with “History-friendly” simulations. Structural Change
and Economic Dynamics 21(4):258-275.
Garud R, Karnøe, P (2001) Path Creation as a Process of Mindful Deviation. In Garud R, Karnøe, P (Org.)
Path Dependence and Creation. Mahwah: Lawrence Erlbaum Associates.
Giddens, A (1984) The Constitution of Society. Berkeley: Univ. California Press.
Hardy C, Maguire, S (2008) Institutional Entrepreneurship. In Greenwood, R et al. (Org.) Handbook of
Organizational Institutionalism. Thousand Oaks: Sage.
Hodgson, GM (1988) Economics and Institutions. Philadelphia: Univ. Pennsylvania Press.
Hodgson GM, Knudsen, T (2004) The complex evolution of a simple traffic convention: the functions and
implications of habit. Journal of Economic Behavior & Organization 54(1):19-47.
Holland, JH (1988) The Global Economy as an Adaptive Process. In Anderson PW, Arrow KJ, Pines, D
(Ed.) The Economy as an Evolving Complex System. Reading: Addison-Wesley.
Hwang H, Powell, WW (2005) Institutions and Entrepreneurship. In Acs Z, Audrestsch, D (Org.) Handbook
of Entrepreneurship Research. New York: Springer.
ITU (2011) Measuring the Information Society: 2011. Geneva: ITU.
Jepperson, R (1991) Institutions, Institutional Effects, and Institutionalization. In Powell WW, DiMaggio, PJ
(Ed.) The New Institutionalism in Organizational Analysis. Chicago: Univ. Chicago Press.
Jonard N, Yildizoğlu, M (1998) Technological diversity in an evolutionary industry model with localized
learning and network externalities. Structural Change and Economic Dynamics 9(1):35-53.
Katz ML, Shapiro, C (1985) Network Externalities, Competition, and Compatibility. American Economic
Review 75(3):424-440.
Kavassalis P, Solomon RJ, Benghozi, PJ (1996) The Internet: a paradigmatic rupture in cumulative telecom
evolution. Industrial and Corporate Change 5(4):1097-1126.
Kirman, A (1997) The economy as an evolving network. Journal of Evolutionary Economics 7(3):339-353.
Klepper, S (1996) Entry, exit, growth, and innovation over the product life cycle. American Economic
Review 86(3):562-583.
Malerba, F (2006) Innovation and the evolution of industries. Journal of Evolutionary Economics 16(1-2):3-
23.
______ (2005) Sectoral Systems of Innovation: a framework for linking innovation to the knowledge base,
structure and dynamics of sectors. Economics of Innovation and New Technology 14(1-2):63-82.
Malerba F, Nelson RR, Orsenigo L, Winter, S (1999) ‘History-friendly’ Models of Industry Evolution: The
Computer Industry. Industrial and Corporate Change 8(1):3-40.
Malerba F, Orsenigo, L (2000) Knowledge, Innovative Activities and Industrial Evolution. Industrial and
Corporate Change 9(2):289-314.
Metcalfe, S (1998) Evolutionary Economics and Creative Destruction. New York: Routledge.
Metcalfe S, Foster, J (2004) Introduction and overview. In Metcalfe S, Foster, J (Ed.) Evolution and
Economic Complexity. Cheltenham: Edward Elgar.
Mowery DC, Simcoe, T (2002) Is the Internet a US invention?: an economic and technological history of
computer networking. Research Policy 31(8-9):1369-1387.
Nelson, RR (1995) Recent Evolutionary Theorizing About Economic Change, Journal of Economic
Literature 33(1):48-90.
Nelson RR, Sampat, B (2001) Making sense of institutions as a factor shaping economic performance.
Journal of Economic Behavior and Organization 44:31-54.
Nelson RR, Winter, SG (1982) An Evolutionary Theory of Economic Change. Cambridge: Harvard Press.
Noam, EM (1994) Beyond liberalisation: From the network of networks to the system of systems”,
Telecommunications Policy 18(4):286-294.
North, DC (1990) Institutions, Institutional Change and Economic Performance. Cambridge: Cambridge
Press.
OECD (2010) Price ranges, Monthly subscriptions, with/without line charge.
www.oecd.org/sti/ict/broadband.
Pavitt, K (1984) Sectoral patterns of technological change: Towards a taxonomy and a theory. Research
Policy 13(6):343-374.
Pereira, MC (2012) O setor de internet no Brasil: uma análise da competição no mercado de acesso.
Dissertação de mestrado, Universidade Estadual de Campinas, Campinas.
Potts, J (2000) New Evolutionary Microeconomics: Complexity, Competence and Adaptive Behaviour.
Cheltenham: Edward Elgar.
Powell, WW (1991) Expanding the scope of institutional analysis. In Powell WW, DiMaggio, PJ (Ed.) The
New Institutionalism in Organizational Analysis. Chicago: Univ. Chicago Press.
Pyka A, Fagiolo, G (2005) Agent-Based Modelling: A Methodology for Neo-Schumpeterian Economics.
Working Paper 272, Institut for Volkswirtschaftslehre, Universitaet Augsburg.
Schumpeter, JA (1943) Capitalism, socialism and democracy. New York: Routledge.
Scott, WR (2008) Institutions and Organizations, 3 ed. Thousand Oaks: Sage.
Simon, HA (1979) Rational decision making in business organizations. American Economic Review
69(4):493-513.
Storper M, Salais, R (1997) Worlds of Production, Cambridge: Harvard Press.
Teece DJ, Pisano G, Shuen, A (1997) Dynamic capabilities and strategic management. Strategic
Management Journal 18(7):509-533.
Tesfatsion, L (2006) Agent-Based Computational Economics: A Constructive Approach to Economic
Theory. In Tesfatsion L, Judd K. (Ed.) Handbook of Computational Economics, Vol. 2: Agent-Based
Computational Economics. Amsterdam: North-Holland.
Thornton P, Ocasio, W (2008) Institutional Logics. In Greenwood, R et al. (Org.) The Sage Handbook of
Organizational Institutionalism, Thousand Oaks: Sage.
Tirole, J (1988) The theory of industrial organization. Cambridge, MA: MIT Press.
Tolbert PS, Zucker, LG (1996) The Institutionalization of Institutional Theory. In Clegg SR, Hardy C, Nord,
WR (Ed.) Handbook of Organization Studies, Thousand Oaks: Sage.
UNCTAD (2010) Information Economy Report 2010: ICTs, Enterprises and Poverty Alleviation. New York:
United Nations.
Valente, M (2009) Markets for Heterogeneous Products: a Boundedly Rational Consumer Model. LEM
Working Paper Series 2009/11.
______ (2002) Simulation Methodology: an Example in Modeling Demand. Mimeo.
Watts, DJ (1999) Networks, Dynamics, and the Small-World Phenomenon. American Journal of Sociology
105(2):493-527.
Williamson, O (2000) The New Institutional Economics: Taking Stock, Looking Ahead. Journal of
Economic Literature 38(3):595-613.
Windrum P, Fagiolo G, Moneta, A (2007) Empirical Validation of Agent-Based Models: Alternatives and
Prospects. Journal of Artificial Societies and Social Simulation 10(2):8-37.